An AI-Driven Predictive Maintenance Framework for Rolling-Element Bearing Fault Diagnosis Using Vibration Signal Analysis

Rolling-element bearings are a major source of failure in rotating machinery, contributing significantly to industrial downtime. This work presents a lightweight, interpretable, and deployable AI-driven predictive maintenance framework for bearing fault diagnosis using vibration signal analysis.

The proposed approach integrates minimal signal pre-processing, computationally efficient time-domain statistical feature extraction, and a Random Forest classifier to achieve robust fault classification under varying operating conditions. Unlike complex deep learning methods, the framework emphasizes simplicity, interpretability, and real-time deployment capability in resource-constrained environments.

Experimental evaluation on the Case Western Reserve University (CWRU) dataset demonstrates an overall classification accuracy of 98.7% and a macro-averaged F1-score of 0.990. Cross-load generalisation achieves 96.2% accuracy on unseen conditions, validating robustness. Statistical validation using McNemar’s test confirms the significance of performance improvements.

Additionally, the trained model is integrated into a web-based diagnostic platform enabling real-time fault prediction, signal visualization, and automated report generation. The results highlight that high diagnostic performance can be achieved using computationally efficient and interpretable methods, making the framework suitable for practical industrial deployment in Industry 4.0 environments.